Hart et al.: Estimates of Streaked abundance in the Republic of Korea 109 ESTIMATING LEUCOMELAS ABUNDANCE IN THE REPUBLIC OF KOREA USING AUTOMATED ACOUSTIC RECORDERS

KIRK A. HART1, STEFFEN OPPEL2, GRANT R.W. HUMPHRIES3, ANDREW BLACKBURN1 & KI-BAEK NAM4*

1Science Faculty, Coast Mountain College, 353 5th Avenue, Prince Rupert, BC V8J 3L6, Canada 2RSPB Centre for Conservation Science, Royal Society for the Protection of , The David Attenborough Building, Cambridge, United Kingdom 3Black Bawks Data Science, Fort Augustus, United Kingdom 4Korea Institute of Ornithology and Department of Biology, Kyung Hee University, Seoul 02447, Republic of Korea *([email protected])

Received 21 July 2020, accepted 28 December 2020

ABSTRACT

HART, K.A., OPPEL, S., HUMPHRIES, G.R.W., BLACKBURN, A. & NAM, K.-B. 2021. Estimating Streaked Shearwater Calonectris leucomelas abundance in the Republic of Korea using automated acoustic recorders. Marine Ornithology 49: 109–117.

Traditional methods to assess population size in are applicable only to that nest in visible locations, leaving cryptic nocturnal, burrow-nesting seabirds underrepresented in typical long-term population monitoring programs. Alternative methods to count burrow-nesting birds, however, are extremely labor-intensive and therefore are applicable only to small and possibly unrepresentative areas, and may have negative effects on burrow-nesting populations. We suggest the use of automatic acoustic recorders as a possible survey technique to overcome logistic difficulties of research of seabirds on remote islands in the Republic of Korea, where funding for ecological research is extremely limited but where globally important colonies exist. In this study, we used automated acoustic recorders to model the relationship between call activity and known abundance of the Streaked Shearwater Calonectris leucomelas at sites on Sasu Island, the largest breeding colony that exists in Korea, supplemented by data from the closely related Cory’s Shearwater C. borealis. Based on a positive relationship between breeding burrow density and call activity, we cautiously estimated the population size of the Streaked to be 95–278 pairs at another breeding colony on Chilbal Island. Although this method may not be precise enough to estimate the exact population size of a seabird species at a location, it can provide coarse estimates that can be used to track relative changes over time.

Key words: automated acoustic recorder, breeding abundance, call activity, Calonectris leucomelas, population estimate, Streaked Shearwater, Korea, passive acoustic monitoring

INTRODUCTION times of day or for months at a time at remote, difficult-to-access sites, such as remote islands. They are particularly useful where Traditional methods to assess population size and trends in nocturnal, burrow-nesting seabirds breed, by utilizing a passive seabirds include point counts and area visual searches (Walsh et al. means of recording (e.g., Buxton et al. 2013, Oppel et al. 2014). In 1995). These methods are applicable only in species whose nests this passive manner, the automated technology avoids the negative are visible, leaving cryptic, nocturnal, burrow-nesting seabirds impacts on wildlife reported in traditional surveying methods and underrepresented in typical long-term monitoring programs facilitates a comprehensive and possibly more representative survey (Schumann et al. 2013). In order to assess the status of the coverage, with less physical effort in the field for the burrow-nesting latter seabirds, some studies have involved burrow searches and populations. Devices can be installed in any area of interest and can nighttime vocalization playback combined with mist netting. These be pre-programmed to record at customized intervals. This reduces methods, however, are extremely labor-intensive and, therefore, are the cost of sampling because no field crew is required during the applicable to small and possibly unrepresentative areas, and they recording period (Williams et al. 2010, Buxton & Jones 2012, may have negative effects on burrow-nesting populations (e.g., Buxton et al. 2013). Importantly, these automated sensors also Bowman et al. 1994, Rodway et al. 1996, Carey 2009). allow for simultaneous, complementary surveying at multiple sites (Borker et al. 2014). Therefore, once the biases and assumptions are In recent years, technological innovations have given researchers quantified, the use of automated acoustic recorders is a cost-effective an alternative method to assess the presence and population size and practical method of monitoring burrow-nesting seabird colonies. of nocturnal, burrow-nesting seabirds: acoustic recorders. The use of automated acoustic recorders is becoming more common in The Streaked Shearwater Calonectris leucomelas is a medium-sized assessments of numbers of marine mammals (Johnson et al. 2009), seabird in the order , with most breeding sites anurans (Brauer et al. 2016), and birds (Buxton & Jones 2012, Buxton located in eastern Asia. Although obtaining quantified information et al. 2013, Borker et al. 2014, Zwart et al. 2014, Pérez-Granados et has been problematic, this species may be decreasing in population al. 2019). These devices are weatherproof sound recording devices size through its entire range in response to the introduction of alien that can be programmed to record bird sounds at pre-determined mammals to breeding islands (Oka 2004, Jones et al. 2008, Dias et

Marine Ornithology 49: 109–117 (2021) 110 Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea al. 2019). Indeed, Norway Rattus norvegicus in the Republic populations over time. To do this, we developed a model that can be of Korea (hereafter Korea) affect the survival and reproductive used to estimate the breeding abundance of Streaked Shearwaters on success of island-nesting seabirds. For example, on Sasu Island, a islands of unknown status using automated acoustic recorders. To small, forested island off the south coast, rats depredate upwards of this end, we first determined the relationship between vocal activity 80% of eggs and chicks in some years (Lee & Yoo 2002, Nam et al. and breeding Streaked Shearwater abundance via quantifying the 2004, Nam et al. 2014, Nam & Yoo 2017). It has become clear that call activity at sites of known density of active burrows. We then eradication projects are a powerful way to improve the conservation used this relationship to estimate abundance on a remote and status of affected seabird species (Howald et al. 2007, Jones et al. inaccessible island for which no information was available about 2016, Brooke et al. 2018). However, to justify the expenditure of Streaked Shearwater population size. funds for an eradication project, knowledge of the seabird breeding populations that exist on these islands, and the ability to monitor METHODS their numbers, must be in place (Holmes et al. 2019). In this regard, the population sizes of Streaked Shearwaters on the Korean islands Study species are almost completely unknown. The Streaked Shearwater breeds on the islands of eastern Asia, Here, we discuss the application of a cost-effective technique that predominantly off Japan (Ochi et al. 2010, Yamamoto et al. 2010, will allow the simultaneous and continuous monitoring of Streaked Sugawa et al. 2014), with smaller colonies around the coast of Shearwater colonies on remote Korean islands, providing a means Korea (Kuroda 1923, Park & Won 1993, Hart et al. 2015) on the to estimate their breeding bird abundance and assess changes in Chinese island of Qingdao in the northern Yellow Sea (Cui 1994),

Bigeum-myeon Republic of Korea Chilbal Island

Docho-myeon

Fig . 1 . Locations of study islands in the Republic of Korea. Automated acoustic recorders were placed on two islands located in southwestern Korea.

Chuja Island Sasu Island

Marine Ornithology 49: 109–117 (2021) Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea 111 on the Pescadores Islands off Taiwan, and on the eastern Russian which is ideal for shearwater burrowing. The dominant seabird island of Karamzina (Oka 2004). species is the Streaked Shearwater, thought to number around 16 000 breeding pairs (KBN, unpubl. data). Non-native Norway Body size and vocalizations of male and female Streaked Shearwaters rats are present and are known to play a significant role in Streaked are sexually dimorphic. Males are generally larger and have a high- Shearwater egg failures (Nam et al. 2014). pitched whistle-like call, while females are slightly smaller and have low-pitched “honking” calls (Shirai et al. 2013, Arima et al. 2014). Chilbal Island (34°47ʹ24ʺN, 125°47ʹ24ʺE), 0.036 km2 in area, is They arrive at their breeding grounds in early March, the first to arrive located ~50 km from the southwestern coast of Korea and is covered awaiting the arrival of their mate; they generally return to the same predominantly by the sedge Carex boottiana along with sparse breeding burrow as the previous year (KBN unpubl. data). Burrows woody vegetation (Park & Won 1993; Figs. 1, 2B). Its highest range from 50 cm to > 2 m in depth and are usually dug into soft soil elevation is 105 m, and its shoreline consists almost entirely of steep of a hillside (Lee et al. 2002). In mid to late June, females lay a single cliffs, amongst which many seabirds nest. There are no permanent egg (Nam et al. 2008), which is incubated until hatching around mid- human residents on the island; however, there is a lighthouse, August; subsequently, the chick fledges in late October (Nam et al. and people visit the island infrequently. Chilbal Island is free of 2014). Adults generally cease visitation about two weeks prior to the introduced mammals and is therefore an important breeding site for fledging of their independent chicks. Swinhoe’s Storm Hydrobates monorhis, of which there are 7 000–13 000 breeding pairs (Taoka et al. 1989, Lee 2010). Streaked Study sites Shearwaters breed on the island, but their abundance is unknown because of the inaccessible nature of the island’s topography. Sasu Island (33°55ʹ12ʺN, 126°37ʹ48ʺE), 0.138 km2 in area, is located between the Korean Peninsula and Jeju Island; it is forested Acoustic recording and deployment protocol and has a peak elevation of ~80 m (Figs. 1, 2A). The island’s perimeter consists mainly of steep cliffs while the interior slopes We recorded the call activity of Streaked Shearwaters using gently, with hills covered in soft soil created from fallen leaf litter, automated acoustic recorders deployed on Sasu Island during the

A

B

Fig . 2 . The typical habitat seen at Sasu (A) and Chilbal (B) islands. Sasu Island is densely forested with soft soil, while Chilbal Island is made up of steep, granite cliffs and is dominated by the grass Carex boottiana.

Marine Ornithology 49: 109–117 (2021) 112 Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea

2014 breeding season (where Streaked Shearwater abundance was Chilbal as compared to recordings from Sasu. We found no major known), and on Chilbal Island during the 2015 breeding season differences in the ability to capture sound between the two islands. (where shearwater abundance was to be estimated). Procellariiform seabirds attend nests mostly continuously during We used Songmeter SM2+ (Wildlife Acoustics Inc., Maynard, MA, the incubation period, with mates trading duties (Ojowski et al. USA), also used in several other related projects (i.e., Borker et al. 2001). After hatching, nest attendance of adult Streaked Shearwaters 2014; Oppel et al. 2014). These devices are waterproof and their becomes less predictable (Ochi et al. 2010), and call activity during recording schedule is customizable, thus making them perfect for this time may not accurately reflect that of the true population size. use on remote seabird islands. The units can accurately capture Therefore, to ensure the highest and most accurate colony attendance vocalizations up to 50 m from the device (Buxton et al. 2013, Oppel for the purpose of modeling, recorders on Sasu and Chilbal islands et al. 2014). were set on the same nightly recording schedule during the incubation period (except plot E, see below). Recording on Sasu and Chilbal Songmeters were placed at four locations, of differing shearwater islands commenced at 23h00 on a 1-min on, 9-min off schedule and density, at least 150 m apart on Sasu Island on 27 June 2014 stopped at 03h00 the following day, giving 24 1-min recordings per (Table 1). The substrate was excellent for burrowing in two plots (A, night. Plot E on Chilbal Island recorded from 23h00 to 03h00 but on B), but at the other two plots (C, D) the shallow soil was mixed with a 2-min on, 28-min off schedule. Recording for Sasu Island started rocks, which limited the number of burrows. The sites were chosen on 24 June 2014 and ended on 31 July 2014 (n = 152 device nights, to represent a range of burrow densities and to cover most of the 60.8 h of recordings). For Chilbal Island during the 2015 breeding breeding habitat on Sasu Island, without overlap of recordings. Due season, recorders operated on slightly different schedules due to to the positioning of the recorders in forests, they were protected technical issues and the remote location of the island. Plot E ran from strong winds which can render recordings inaudible. As a 26 June 2015–02 July 2015 (n = 7 device nights around new moon); result, all recorders produced files that had very clear audio and plot F ran 26 June 2015–20 July 2015 (n = 25 device nights); plot G produced spectrograms from which Streaked Shearwater calls ran 26 June 2015–12 July 2015 and 20 July 2015–31 July 2015 were easily recognizable. No recordings had to be discarded due to (n = 29 device nights). The recording schedule for devices on plots F inaudibility from strong winds or other noise. and G included all stages of a lunar cycle.

On Chilbal Island, single recorders were placed on the south Acoustic data analysis (plot E), west (plot F), and north (plot G) slopes, at least 100 m apart, on 01 June 2015; due to the island’s small size, these three Acoustic data were analyzed using Song Scope 4.1.3A (Wildlife recorders were capable of recording the vocalizations of nearly the Acoustics Inc.). Originally, we attempted to use an automated entire island without overlap (Table 1). Given the lack of trees, we detection algorithm in Song Scope (Buxton et al. 2013, Oppel expected many recordings to be inaudible because of strong winds. et al. 2014), but due to the high overlap of Streaked Shearwater Therefore, we applied soft wind screens to each microphone. In calls at densely inhabited sites on Sasu Island, and the overlap addition, we placed recorders behind large boulders to function as of Streaked Shearwater calls with Swinhoe’s Storm on wind breaks. Consequently, only a small number of recordings on Chilbal Island, the software yielded high rates of false positives fewer than five nights were unusable. and negatives. Therefore, we decided to manually quantify calls by visually observing spectrograms. In some cases, we listened to the On both islands, recorders were placed as far from the ocean as respective audio to ensure correct identification and quantification. possible to reduce wave noise interference. We set the sample rate at 16 kHz and microphones at a gain of +42.0 dB. Devices were We defined a call as any repeating syllable in a Streaked Shearwater attached to plywood bases screwed to a wooden post 1 m in length vocalization. While male and female calls can be easily distinguished, that was inserted in the ground. To provide additional stability, large we counted the total number of calls of both sexes combined in rocks were placed around the base of the post. We tested whether each 1-min recording. Because call activity of nocturnal seabirds we were able to detect experimental noises and human voices at is known to vary widely from night to night due to a number of similar distances and with equal probability in recordings from environmental factors (Granadeiro et al. 2009, Bretagnolle et al.

TABLE 1 Nocturnal call activity of Streaked Shearwaters Calonectris leucomelas at study sites on Sasu and Chilbal islands, monitored using automated acoustic recorders during the 2014 and 2015 breeding seasons, respectively Recording Total recording time Total number Mean number Study site Coordinates plot (min) of calls of calls/min Sasu Island A 33°55ʹ14.7ʺN, 126°38ʹ27.0ʺE 840 93 219 110.975 B 33°55ʹ13.7ʺN, 126°38ʹ22.8ʺE 744 45 940 61.747 C 33°55ʹ12.9ʺN, 126°38ʹ18.6ʺE 836 39 334 47.050 D 33°55ʹ13.5ʺN, 126°38ʹ14.8ʺE 816 21 835 26.759 Chilbal Island E 34°47ʹ15.5ʺN, 125°47ʹ19.1ʺE 42 274 6.523 F 34°47ʹ15.3ʺN, 125°47ʹ14.9ʺE 515 2 646 5.137 G 34°47ʹ18.3ʺN, 125°47ʹ17.0ʺE 760 836 1.100

Marine Ornithology 49: 109–117 (2021) Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea 113

2012), the instantaneous call rate during any given 1-min recording shearwaters are roughly 1–2 s long, strung together in near can vary by orders of magnitude. The main benefit of long-term continuous vocalizations with gaps of 0–1 s between syllables, and automated recording is the regular and consistent recording over appear to vary more between sexes in each species than among the different phases of the incubation cycle, the nocturnal cycle, and two species (Bretagnolle & Lequette 1990, Arima et al. 2014). across moon phases, thus overcoming the known variation in Procellariform attendance patterns and calling rates. Therefore, We averaged the number of calls/min over the entire recording we took an average of all recordings during incubation for each period per recording site. We then related this index of vocal activity site and used the average number of calls per min to calibrate the against Calonectris shearwater nest density using a generalized linear relationship between nest density and vocal activity. model (GLM), with ‘species × calling rate’ as an interaction term to allow for species-specific calling rates; models without species- Breeding burrow assessment specific calling rates were inferior in preliminary explorations (Table 3). This interaction also accounts for the fact that Cory’s Under the assumption that song meter units are capable of recording Shearwater density was lower and was not extrapolated based on identifiable vocalizations up to 50 m away (Buxton et al. 2013, occupancy rate from sampling quadrats. Cory’s Shearwater density Oppel et al. 2014), in order to estimate the number of Streaked was determined by inspecting all available breeding burrows and Shearwaters in the area surrounding each recorder, we searched was fully enumerated in a 50-m radius around recorders (Oppel et for occupied burrows within 50 m of the device on Sasu Island al. 2014). (Table 2). Because searching for all burrows in the entire 50 m radius was logistically infeasible, we established 10 randomly With the average nocturnal call rate for Chilbal Island known for dispersed sampling quadrats (dimension: 10 × 10 m) within the each site, we applied the model produced in the previous step to 50-m radius around each recorder. Streaked Shearwater burrows these data and estimated the number of breeding burrows at each of regularly exceed 2 m in length and often require excavation to the three recording sites. We produced an estimate along with 95% check status, an effort that is time-consuming and stressful for confidence intervals, and all statistical analyses were carried out occupants. This, coupled with limits to our permitted access to Sasu with R version 4.0.0 (R Core Team 2020). Island, necessitated that we use an active breeding pair occupancy rate derived from previous work of 82.6% (Nam et al. 2014). We RESULTS averaged the estimated number of breeding burrows/m2 across all 10 quadrats and multiplied this average density by the total area of Nocturnal call activity each recording zone (7 853.98 m2) to obtain an estimate of actively breeding Streaked Shearwaters around each recording unit. In total, 3 236 1-min recordings were collected at the four recording sites on Sasu Island and a total of 200 328 Streaked Shearwater Estimating the breeding population size calls were identified. The number of calls per min ranged 0–357, and the average number of calls varied 26.759–110.975 among the To estimate breeding bird abundance on Chilbal Island based on four sites (Table 1). calling activity, we needed to understand the relationship between call activity and abundance at each recorder site on Sasu Island. On Chilbal Island, a total of 1 317 min of recordings were collected In order to add to the four recording sites on Sasu Island and and a total of 3 756 Streaked Shearwater calls were identified across make our model more robust, we included information on the the three recording sites. The number of calls per min ranged 0–50, call activity and number of nests in 11 plots in colonies of the and the average number of calls at each site ranged 1.100–6.523 closely related Cory’s Shearwater Calonectris borealis among the (Table 1). Azores Islands, using the same recording schedule as Sasu Island (Oppel et al. 2014; Appendix 1, available on the website). Both Breeding burrow density species display behavioural similarities, including comparable call behaviour (Bretagnolle & Lequette 1990, Rabouam et al. The number of burrows in each of the 10 sample quadrats varied 2000, Arima et al. 2014). The calls of both Cory’s and Streaked very little within each of the four recording sites on Sasu Island,

TABLE 2 The estimation of active burrows of Streaked Shearwaters Calonectris leucomelas at four recording plots on Sasu Island, Republic of Korea, during the incubation period, 2014 Total number of Number of burrows in 10 Number of estimated active Active burrow density Plota estimated active burrows quadratsb burrowsc in 10 quadrats (No . of burrows/m2 ± SD) in a recording plot A 600 496 0.50 ± 0.08 3 927 B 385 318 0.32 ± 0.11 2 513 C 106 88 0.09 ± 0.05 707 D 140 116 0.12 ± 0.09 942 a Each plot area was 7 853.98m2 b All quadrats were 10 × 10 m c A burrow occupancy rate of 0.826 was applied to estimated total burrows to estimate active burrows.

Marine Ornithology 49: 109–117 (2021) 114 Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea

TABLE 3 Candidate models describing breeding bird abundance (number of active breeding burrows) of Streaked Shearwaters Calonectris leucomelas in relation to the average nightly call activity (Mean_calls_min) and Calonectris species . Model 1 Model 2 Model 3a Incidence Confidence Incidence Confidence Incidence Confidence Predictors P P P rate ratios interval rate ratios interval rate ratios interval (Intercept) 1.03 0.90–1.19 0.637 2.14 1.81–2.52 <0 .001 14.66 11.09–18.88 <0 .001 Mean_calls_min [log] 5.90 5.71–6.10 <0 .001 3.08 2.94–3.22 <0 .001 1.44 1.30–1.61 <0 .001 Species [Streaked Shearwater] 8.89 7.81–10.15 <0 .001 0.91 0.66–1.29 0.604 Mean_calls_min [log] × 2.32 2.06–2.60 <0 .001 Species [Streaked Shearwater] R2 0.89 0.95 0.96 Akaike Information Criterion 2 454.5 1 116.4 981.8 a Model 3 was used to estimate breeding bird abundance on Chilbal Island. but there were large differences across sites (Table 2). Plot A had recorded there was higher than it would have been if averaged over the highest number of burrows within the quadrats and, thus, the the full lunar cycle. Therefore, the higher average mean calling highest mean density of breeding burrows (0.50 ± 0.08 nests/m2), rate extrapolated from plot E may overestimate the abundance of followed by plot B (0.32 ± 0.11 nests/m2). Substantially lower Streaked Shearwaters. density existed in the remaining two plots (Table 2). DISCUSSION Estimating breeding population size Chilbal Island is a breeding colony primarily for Swinhoe’s We found a positive relationship between average nocturnal call Storm Petrels (Park & Won 1993, Lee 2010), as well as a small rate of Calonectris shearwaters and the number of active breeding number of Streaked Shearwaters. Based on acoustic recordings, we burrows within a 50-m radius (R2 = 0.96, P < 0.001; Fig. 3). By cautiously suggest that 90–280 pairs of Streaked Shearwaters may applying the fitted regression equation to call activity data from nest on this island, but this estimate is surrounded by considerable Chilbal Island, we estimated the Streaked Shearwater population on uncertainty. Nonetheless, our estimates of small numbers of Chilbal Island (Table 4). Average call activity ranged 0–50 calls per Streaked Shearwaters are consistent with anecdotal accounts, which min and resulted in a total estimate of 208–278 Streaked Shearwater are solely based on opportunistic observations without detailed pairs (95–13 pairs excluding plot E) breeding on Chilbal Island surveys (Kuroda 1923, Kang et al. 2008). Given the automated in 2015 (Table 4). Call activity at plots F and G decreased as the and repeatable approach to estimation that we provide, our data recording period progressed, coinciding with increasing moon can be combined with future estimates for monitoring relative illumination. Because the recorder at plot E operated only during changes over time at small- and medium-sized colonies, where the dark, new phase of the moon, it is likely that the call activity oversaturation of recorders is not an issue.

Although Sasu Island is only about three times larger, it had

Species 20–100 more calls/min and a larger nest density than Chilbal Cory’s Shearwater Island. This is simply due to the limited nesting habitat on Chilbal 9 Streaked Shearwater Island, much of which is covered by the grass Carex boottiana. The roots of this grass make burrowing very difficult for Streaked Shearwaters, thus reducing the availability of suitable burrowing 6 soil. On the other hand, the much smaller Swinhoe’s Storm Petrel

3 TABLE 4 Estimated number of breeding burrows of log(Number of breeding burrows) log(Number of breeding Streaked Shearwaters Calonectris leucomelas on 2 3 4 5 log(Mean calls/minute) three recording sites at Chilbal Island based on call activity Estimated number Fig . 3 . The relationship between breeding burrow abundance and Recording Mean number of breeding burrows the calling rate of two Calonectris species. The total number of plot of calls/min active burrows within a 50-m radius of the automated acoustic (95% Confidence Interval) recorders is represented by breeding burrows in this figure (solid E 6.523 129 (113–147) line = fitted regression line based on a generalised linear model F 5.137 96 (83–112) with a species × calling rate interaction term; shaded area = 95% confidence intervals). G 1.100 15 (12–19)

Marine Ornithology 49: 109–117 (2021) Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea 115 can more easily navigate the roots to produce a burrow, making it 2015), would serve as ideal sites, as they host relatively smaller, the dominant species on Chilbal Island. easily accessible populations.

Ideally, the quantification of data captured by automated acoustic Our goal in this study was to establish an automated technique recorders would use algorithms in an entirely automatic process. for efficiently and effectively surveying a nocturnal, burrow- This would significantly reduce processing time and allow for nesting seabird on islands around the Korean Peninsula, especially thousands of hours of recordings to be captured and analyzed, those on which inaccessible terrain prevents traditional surveying thus providing the most accurate estimate of island call activity. methods. Our approach provided the first full-island survey of While there has been a lot of progress in the effectiveness of Streaked Shearwaters on Chilbal Island, and its application seems such algorithms since the time of this study (Stowell et al. 2019, promising. This method may not precisely estimate the population Priyadarshani et al. 2020, Wright et al. 2020), most studies deal size breeding on an island, but it offers a repeatable method that can only with small numbers of calling individuals within the area be used to quantify relative changes in species numbers over time of the recorders and thus do not face the problem of overlapping in response to introduced species removal or other factors. When calls, which complicates automatic detection and enumeration considering that nearly all Korean Streaked Shearwater colonies (Orben et al. 2019, Arneill et al. 2020). On Sasu Island, however, of unknown status exist on small islands with no other nocturnally the shearwater call rates at some sites are > 100 per min, which vocalizing species (Hart et al. 2015), we believe these automated leads to significant overlap of calls. This causes high rates of acoustic recorders and associated recognition software could work false positives and negatives in the spectrogram analysis of some well as a low-cost, user-friendly survey technique. Their use would automatic detection software (Buxton & Jones 2012). Similarly, fill in the vital information gaps at these remote areas for Streaked because of the high rate of Swinhoe’s Storm Petrel calls on Chilbal Shearwaters and other nocturnal, burrow-nesting seabirds. Island, it is possible that the calls of a different species could mask the calls of Streaked Shearwaters, causing Streaked Shearwater ACKNOWLEDGEMENTS calls to be underestimated. These two factors, unfortunately, prevent the use of automatic detection software in its current form We thank members of the Ecology Laboratory in Kyung on Sasu and Chilbal islands. Hee University for their assistance with field work. We also thank Alex Bond, Chang-Yong Choi, and Ian Jones for providing valuable We presumed that the vocalizations that we quantified reflected information and discussion, and Miran Kim for assistance on the breeding population of Streaked Shearwaters in this area based Chilbal Island. This research was supported by the Oriental Bird on our assumption that the proportion of vocalizations coming Club through the AEC Award, IdeaWild, and the Basic Science from non-breeding individuals was similar across sites. Because Research Program through the National Research Foundation non-breeding individuals in some shearwaters have higher levels of Korea (NRF), funded by the Ministry of Education (NRF- of call activity than breeders (James 1985, Arneill et al. 2020), 2013R1A1A2058587). Comments by anonymous reviewers helped we acknowledge that if a larger proportion of non-breeders was us to greatly improve the paper. present on Chilbal, or if there was high non-breeder movement around the colony (Arneill et al. 2020), these birds may have REFERENCES inflated our assessment of population size on Chilbal Island. Additionally, we assumed that the difference in surface vegetation ARIMA, H., OKA, N., BABA, Y., SUGAWA, H. & OTA, T. 2014. (Sasu = forested; Chilbal = grass-only) would not affect the audio Gender identification by calls and body size of the Streaked capturing ability of the recorders. In field tests, only heavy rain Shearwater examined by CHD genes. Ornithological Science and wind were found to alter audio recording quality within 13: 9–17. doi:10.2326/osj.13.9 the 50-m recording zone, and both islands had nearly identical ARNEILL, G.E., CRITCHLEY, E.J., WISCHNEWSKI, S., detection rates for experimental recordings. JESSOPP, M.J. & QUINN, J.L. 2020. Acoustic activity across a seabird colony reflects patterns of within-colony flight rather We acknowledge that our small sample size is a limiting factor than nest density. Ibis 162: 416–428. doi:10.1111/ibi.12740 for extrapolating our results, and that the small sample size BORKER, A.L., MCKOWN, M.W., ACKERMAN, J.T., EAGLES- and resulting uncertainty are therefore the key weakness of the SMITH, C.A., TERSHY, B.R. & CROLL, D.A. 2014. Vocal study. To overcome the small sample size of Korean colonies, activity as a low cost and scalable index of seabird colony size. we included data from a closely related species. However, Conservation Biology 28: 1100–1108. doi:10.1111/cobi.12264 these two species’ datasets did not overlap in either call activity BOWMAN, T., THOMPSON, S., JANIK, C. & DUBUC, L. or number of breeding burrows and were based on a slightly 1994. Nightlighting minimizes investigator disturbance in bird different assessment of nest abundance around sound recorders. colonies. Colonial Waterbirds 17: 78–82. Hence, we were unable to ascertain whether the differences we BRAUER, C.L., DONOVAN, T.M., MICKEY, R.M., KATZ, J. & found between the two studies were solely due to population MITCHELL, B.R. 2016. A comparison of acoustic monitoring size or whether other confounding aspects may have affected methods for common anurans of the northeastern United States. vocalization characteristics at low densities (Fig. 3). Therefore, Wildlife Society Bulletin 40: 140–149. doi:10.1002/wsb.619 our model would benefit from the inclusion of additional Streaked BRETAGNOLLE, V., GENEVOIS, F. & MOUGEOT, F. 2012. Shearwater colonies with small, known populations to allow for Intra- and intersexual functions in the call of a non-passerine overlap of those variables and thus allow us to better assess model bird. Behaviour 135: 1161–1184. quality. Given that Chilbal Island is logistically challenging and, BRETAGNOLLE, V. & LEQUETTE, B. 1990. Structural therefore, we cannot perform a physical population estimate to variation in the call of the Cory’s Shearwater (Calonectris validate our predictions, other breeding colonies of Streaked diomedea, Aves, ). Ethology 85: 313–323. Shearwaters, such as Hwa Island and Gugul Island (Hart et al. doi:10.1111/j.1439-0310.1990.tb00410.x

Marine Ornithology 49: 109–117 (2021) 116 Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea

BROOKE, M. de L., BONNAUD, E., DILLEY, B.J. ET AL. 2018. LEE, K.-G., NAM, K.-B., LEE, J.-Y., KIM, H.-J. & YOO, D.S. Seabird population changes following mammal eradications on 2002. Morphological characteristics of burrows, adults and islands. Animal Conservation 21: 3–12. doi:10.1111/acv.12344 eggs of Streaked Shearwaters (Calonectris leucomelas). Korean BUXTON, R.T. & JONES, I.L. 2012. Measuring nocturnal seabird Journal of Ornithology 9: 23–29. activity and status using acoustic recording devices: applications LEE, K.-G. & YOO, J.-C. 2002. Breeding population of Streaked for island restoration. Journal of Field Ornithology 83: 47–60. Shearwaters (Calonectris leucomelas) and the effect of Norway doi:10.1111/j.1557-9263.2011.00355.x (Rattus norvegicus) predation on Sasudo Island. Journal BUXTON, R.T., MAJOR, H.L., JONES, I.L. & WILLIAMS, J.C. 2013. of the Yamashina Institute for Ornithology 33: 142–147. Examining patterns in nocturnal seabird activity and recovery across doi:10.3312/jyio1952.33.142 the Western Aleutian Islands, Alaska, using automated acoustic NAM, K.-B., KWON, I.-K. & YOO, D.S. 2004. Causes of hatching recording. The Auk 130: 331–341. doi:10.1525/auk.2013.12134 failure of Streaked Shearwaters Calonectris leucomelas on CAREY, M.J. 2009. The effects of investigator disturbance on Sasudo Island. Korean Journal of Ornithology 11: 79–85. procellariiform seabirds: a review. New Zealand Journal of NAM, K.-B., KWON, I.K. & YOO, J.C. 2008. Incubation routine Zoology 36: 367–377. doi:10.1080/03014220909510161 and sex role of Streaked Shearwaters Calonectris leucomelas CUI, S. 1994. Study on ecology and migration of White Fronted at Sasudo Island, South Korea. Ocean and Polar Research 30: Shearwater. Chinese Journal of Zoology 3: 24–33. 11–19. doi:10.4217/OPR.2008.30.1.011 DIAS, M.P., MARTIN, R., PEARMAIN, E.J. ET AL. 2019. Threats NAM, K.-B. & YOO, J.C. 2017. Population trend of a colonial to seabirds: a global assessment. Biological Conservation 237: breeding seabird under high predation risk. Proceedings of the 525–537. doi:10.1016/j.biocon.2019.06.033 2017 Fall Conference of the Ornithological Society of Korea. GRANADEIRO, J.P., ALONSO, H., ALMADA, V., MENEZES, Busan, South Korea: Ornithological Society of Korea. D., PHILLIPS, R.A. & CATRY, P. 2009. Mysterious attendance NAM, K.-B. & YOO, J.C. 2017. Population trend of a colonial cycles in Cory’s shearwater, Calonectris diomedea: an exploration breeding seabird, under high predation risk. Ornithological of patterns and hypotheses. Animal Behaviour 78: 1455–1462. Society of Korea, Proceedings of 2017 Fall Conference. doi:10.1016/j.anbehav.2009.09.029 OCHI, D., OKA, N. & WATANUKI, Y. 2010. Foraging trip HART, K.A., CHOI, C.-Y., BOND, A.L., HUMPHRIES, G.R.W., decisions by the streaked shearwater Calonectris leucomelas YOO, J.-C. & NAM, K.-B. 2015. Streaked Shearwaters depend on both parental and chick state. Journal of Ethology 28: Calonectris leucomelas of the Korean Peninsula: distribution, 313–321. doi:10.1007/s10164-009-0187-3 status and potential threats. Forktail 31: 55–63. OJOWSKI, U., EIDTMANN, C., FURNESS, R. & GARTHE, S. HOLMES, N.D., SPATZ, D.R., OPPEL, S. ET AL. 2019. Globally 2001. Diet and nest attendance of incubating and chick-rearing important islands where eradicating invasive mammals will northern fulmars (Fulmarus glacialis) in Shetland. Marine benefit highly threatened vertebrates. PLoS One 14: 1–17. Biology 139: 1193–1200. doi:10.1007/s002270100655 doi:10.1371/journal.pone.0212128 OKA, N. 2004. The distribution of Streaked Shearwater colonies, HOWALD, G., DONLAN, C.J., GALVÁN, J.P. ET AL. 2007. with special attention to population size, area of sea where Invasive rodent eradication on islands. Conservation Biology 21: located and surface water temperature. Journal of the Yamashina 1258–1268. doi:10.1111/j.1523-1739.2007.00755.x Institute of Ornithology 35: 164–188. JAMES, P.C. 1985. The vocal behaviour of the Manx Shearwater OPPEL, S., HERVÍAS, S., OLIVEIRA, N. ET AL. 2014. Estimating puffinus. Zeitschrift für Tierpsychologie 67: 269–283. population size of a nocturnal burrow-nesting seabird using doi:10.1111/j.1439-0310.1985.tb01394.x acoustic monitoring and habitat mapping. Nature Conservation JOHNSON, M., DE SOTO, N.A. & MADSEN, P.T. 2009. Studying 7: 1–13. doi:10.3897/natureconservation.7.6890 the behaviour and sensory ecology of marine mammals using ORBEN, R.A., FLEISHMAN, A.B., BORKER, A.L. ET AL. 2019. acoustic recording tags: a review. Marine Ecology Progress Series Comparing imaging, acoustics, and radar to monitor Leach’s 395: 55–73. doi:10.3354/meps08255 storm-petrel colonies. PeerJ 7: e6271. doi:10.7717/peerj.6721 JONES, H.P., HOLMES, N.D., BUTCHART, S.H.M. ET AL. 2016. PARK, J.Y. & WON, P.O. 1993. Survey of seabirds breeding Invasive mammal eradication on islands results in substantial in Korea. The Bulletin of Korea Institute of Ornithology 4: conservation gains. Proceedings of the National Academy of 101–105. Sciences of the United States of America 113: 4033–4038. PÉREZ-GRANADOS, C., BOTA, G., GIRALT, D. ET AL. 2019. doi:10.1073/pnas.1521179113 Vocal activity rate index: a useful method to infer terrestrial JONES, H.P., TERSHY, B.R., ZAVALETA, E.S. ET AL. 2008. bird abundance with acoustic monitoring. Ibis 161: 901–907. Severity of the effects of invasive rats on seabirds: a global doi:10.1111/ibi.12728 review. Conservation Biology 22: 16–26. doi:10.1111/j.1523- PRIYADARSHANI, N., MARSLAND, S., JUODAKIS, J., 1739.2007.00859.x CASTRO, I. & LISTANTI, V. 2020. Wavelet filters for automated KANG, J., KANG, T., YOO, S., CHO, H., LEE, S. & KIM, I. recognition of birdsong in long-time field recordings. Methods 2008. Study on the breeding status of the natural monument in Ecology and Evolution 11: 403–417. doi:10.1111/2041- islet (chilbaldo, sasudo, nando, hongdo). Korean Journal of 210X.13357 Ornithology 15: 169–175. R CORE TEAM. 2020. R: A language and environment for KURODA, N. 1923. The birds at Nishi Island lighthouse in Hwanghae statistical computing. Vienna, Austria: The R Foundation for Do, Korea. Tori 3: 309–314. Statistical Computing. LEE, K.-G. 2010. The status of seabirds on Sasu and Chilbal RABOUAM, C., BRETAGNOLLE, V., BIGOT, Y. & PERIQUET, islands, and the management of invasive species. The Birds Korea G. 2000. Genetic relationships of Cory’s Shearwater: Blueprint 2010 for the conservation of the avian biodiversity of parentage, mating assortment, and geographic differentiation the South Korean part of the Yellow Sea, pp. 134–137. Busan, revealed by DNA fingerprinting. The Auk 117: 651–662. Republic of Korea: Birds Korea. doi:10.2307/4089591

Marine Ornithology 49: 109–117 (2021) Hart et al.: Estimates of Streaked Shearwater abundance in the Republic of Korea 117

RODWAY, M.S., MONTEVECCHI, W.A. & CHARDINE, WALSH, P.M., HALLEY, D.J., HARRIS, M.P., DEL NEVO, A., J.W. 1996. Effects of investigator disturbance on SIM, I.M.W. & TASKER, M.L. 1995. Seabird monitoring breeding success of Atlantic puffins Fratercula arctica. handbook for Britain and Ireland: a compilation of methods Biological Conservation 76: 311–319. doi:10.1016/0006- for survey and monitoring of breeding seabirds. London, UK: 3207(94)00118-9 Institute of Terrestrial Ecology, and Seabird Group. SCHUMANN, N., DANN, P., HOSKINS, A.J. & ARNOULD, WILLIAMS, J.C., DRUMMOND, B.A. & BUXTON, R.T. 2010. J.P.Y. 2013. Optimizing survey effort for burrow-nesting Initial effects of the August 2008 volcanic eruption on breeding seabirds. Journal of Field Ornithology 84: 69–85. birds and marine mammals at Kasatochi Island, Alaska. Arctic, doi:10.1111/jofo.12007 Antarctic, and Alpine Research 42: 306–314. doi:10.1657/1938- SHIRAI, M., NIIZUMA, Y., TSUCHIYA, K., YAMAMOTO, 4246-42.3.306 M. & OKA, N. 2013. Sexual size dimorphism in Streaked WRIGHT, W.J., IRVINE, K.M., ALMBERG, E.S. & LITT, A.R. Shearwaters Calonectris leucomelas. Ornithological Science 2020. Modelling misclassification in multi-species acoustic 12: 57–62. doi:10.2326/osj.12.57 data when estimating occupancy and relative activity. Methods STOWELL, D., WOOD, M.D., PAMUŁA, H., STYLIANOU, in Ecology and Evolution 11: 71–81. doi:10.1111/2041- Y. & GLOTIN, H. 2019. Automatic acoustic detection of 210X.13315 birds through deep learning: The first bird audio detection YAMAMOTO, T., TAKAHASHI, A., KATSUMATA, N., SATO, challenge. Methods in Ecology and Evolution 10: 368–380. K. & TRATHAN, P.N. 2010. At-sea distribution and behavior doi:10.1111/2041-210X.13103 of Streaked Shearwaters (Calonectris leucomelas) during the SUGAWA, H., KARINO, K., OHSHIRO, A. & HIRAI, M. nonbreeding period. The Auk 127: 871–881. doi:10.1525/ 2014. Long-term trends in breeding site fidelity of streaked auk.2010.10029 shearwater Calonectris leucomelas. Marine Ornithology 42: ZWART, M.C., BAKER, A., MCGOWAN, P.J.K. & 11–15. WHITTINGHAM, M.J. 2014. The use of automated bioacoustic TAOKA, M., WON, P.O. & OKUMURA, H. 1989. Vocal behavior recorders to replace human wildlife surveys: an example of Swinhoe’s Storm-Petrel (Oceanodroma monorhis). The using nightjars. PLoS One 9: 102770. doi:10.1371/journal. Auk 106: 471–474. pone.0102770

Marine Ornithology 49: 109–117 (2021)